According to Chao(1965):
A form[1] in the widest sense is any fraction of a language, be it a word or any succession of words. In practice, however, the smallest units, such as phonemes, features of a phoneme, say voicelessness, or very large units, such as a whole lecture or a play, are not forms as object of grammatical analysis. Only the smallest unit that has meaning, the morpheme, and the largest unit between pauses, the sentence,[2] and units of intermediate sizes have generally been included in grammatical analysis. Forms usually occur in succession, except for a relatively small number of simultaneous or overlapping forms, such as an interrogative intonation over a whole sentence, which can be considered a morpheme having a meaning but residing parasitically on the other forms, the words in the sentence. Forms which are worth considering in grammar have a certain unity such that they will recur with similar forms in similar contexts. Such forms are then constituents, as distinguished from any fraction of a larger form, cut out arbitrarily, which makes up a string but not necessarily a constituent. For example, in 说中国话 shuo Jong. gwohuah ‘speak Chinese’, each of the syllables, which is also a morpheme, is a constituent. Furthermore, shuo, Jong. gwohuah, Jong. gwo, and huah are constituents, but shuo Jong and . gwohuah are not constituents. In other contexts shuo Jong. gwo ‘speaks of China’ could be a constituent, but it is not one in shuo Jong. gwohuah. It goes without saying that a fractional part of a morpheme plus whatever follows or precedes will make up a string but not a constituent.
[1] It is an awkward discrepancy in usage between Chinese and Western writers that Chinese 形 shyng ‘form (of characters)’ is contrasted with 音 in ‘sound’ and 义 yih ‘meaning’, while ‘form’ in Western usage corresponds really to in and not to shyng. The confusion is compounded when recent Chinese writers translate ‘morphology’ as 形态 shyngtay ‘form-state’ and when recent Western writers go further and use ‘shape’ for a particular phonemic make-up of a form. However, one can usually tell what an author means, given the context of use.
[2]A beginning, however, has been made in the formal study of larger units by Zellig S. Harris, in his “Discourse Analysis”, Language, 28:1-30, and “Discourse Analysis: A Sample Text”, Language, 28: 474-494 (1952).
[3]Chao Y R. A grammar of spoken Chinese[M]. Univ of California Press, 1965.